Preamble

Aware that artificial intelligence constitutes one of the major transformations of our time, carrying both considerable progress and unprecedented risks; that its uses may affect human dignity, fundamental freedoms, the security of peoples, the integrity of institutions, economic, social and environmental balances, as well as peace among nations;

Recalling that peace cannot be reduced merely to the absence of conflict, but also presupposes justice, security, cooperation, respect for the rule of law, and the protection of the human person;

Reaffirming the heritage of humanist and universalist traditions, as well as the authority of international legal instruments, in particular the Charter of the United Nations, the Universal Declaration of Human Rights, the International Covenants of 1966, the European Convention on Human Rights, and the Geneva Conventions;

Considering that this Charter forms part of the continuity of the work and dialogues undertaken within the framework of the global forum “Normandy for Peace,” whose reflection on the ethical, legal, and political conditions for artificial intelligence serving peace it seeks to extend;

Believing that the development of artificial intelligence calls for a shared responsibility among States, economic actors, the scientific community, and citizens;

The signatories adopt this Charter as an ethical framework and guiding principle for action, with a view toward a future international legal instrument intended to prevent the extreme risks associated with artificial intelligence and to ensure that it remains in the service of peace.

Article 1 — Primacy of Human Dignity and the Imperative of Peace

Artificial intelligence may neither be conceived, developed, nor used except in full respect for human dignity, fundamental rights, and the imperative of peace.

Any initiative relating to artificial intelligence must be subject to a prior assessment of its impacts on individuals and societies, in order to prevent any harm to physical, psychological, collective, social, or structural integrity.

No consideration of profit, power, or convenience may prevail over the protection of the human person and the preservation of peace.

Article 2 — Effective and Permanent Human Control

Humanity must, under all circumstances, remain in control of artificial intelligence systems.

No system exerting significant influence over individuals, communities, or institutions may be deployed without guaranteeing genuine, informed, continuous, and effective human oversight.

Such control implies the ability to supervise, understand, correct, suspend, or interrupt the functioning of a system.

High-impact systems must incorporate, from the design stage onward, reliable deactivation or emergency shutdown mechanisms that may be activated by competent authorities whenever a serious risk is identified.

Every person must be able to challenge an automated decision affecting them and obtain effective human intervention.

Article 3 — Health and Human Integrity

The use of artificial intelligence in the field of healthcare shall remain subject to the primacy of human medical judgment.

No essential decision regarding diagnosis, treatment, orientation, or triage may be delegated to an automated system without qualified human supervision.

The free and informed consent of the patient constitutes a fundamental condition for any use of artificial intelligence in healthcare, subject to the strictly limited exceptions provided by law.

Health data must benefit from enhanced protection and may not be commercialized in a manner contrary to human dignity.

Medical systems must be traceable, verifiable, intelligible, and free from discriminatory bias.

Access to the benefits of artificial intelligence in healthcare must be ensured equitably, without discrimination based on income, origin, or territory.

Article 4 — Transparency and Traceability

Competent authorities must be able clearly to determine what constitutes an artificial intelligence system under this Charter.

High-impact systems must be documented, traceable, auditable, and explainable to a degree compatible with their nature and uses.

Their purpose, responsible parties, principal data sources, known limitations, and conditions of deployment must be identifiable.

Transparency constitutes an essential condition for public trust, accountability, and democratic oversight.

Article 5 — Combating Manipulation and Disinformation

The use of artificial intelligence for purposes of cognitive manipulation, organized disinformation, or the distortion of legitimate and representative collective decision-making processes is prohibited.

Any content in which artificial intelligence has played a substantial role in creation, alteration, or dissemination must be identifiable through appropriate means.

Digital actors are required to prevent, detect, and limit the mass dissemination of such content, under the supervision of competent authorities and with due respect for fundamental freedoms.

Article 6 — Prohibition of Extreme-Risk Uses

It is prohibited to design, develop, make available, or deploy artificial intelligence systems liable to generate an extreme, irreversible, or global-scale risk to humanity.

This includes, in particular:

• systems presenting a massive or uncontrollable lethal risk;
• systems escaping human control on a lasting basis;
• systems enabling the critical concentration of power over essential global infrastructures in the hands of a single actor;
• systems capable of causing severe systemic harm to societies or ecosystems.

An international classification of risks must make it possible to identify such prohibited uses.

Designers, developers, and operators shall incur civil and criminal liability where damage results from manifestly insufficient safety measures. It shall be their responsibility to demonstrate that all reasonably expected precautions were taken.

Article 7 — Prohibition of Autonomous Militarization and Mass Surveillance

No machine shall be entrusted with the power to decide over life or death.

Any form of lethal autonomy lacking meaningful human intervention is prohibited.

Likewise prohibited are systems of mass surveillance for military or security purposes enabling generalized targeting, predictive profiling, or disproportionate infringements upon fundamental freedoms.

These principles are grounded in respect for international humanitarian law.

Article 8 — National Sovereignty and Global Security

The security of populations, the continuity of essential services, and the sovereignty of States constitute overriding imperatives.

Any artificial intelligence system liable to compromise critical infrastructures — including energy, healthcare, financial, digital, or institutional infrastructures — must be prohibited or strictly regulated.

In the event of serious uncertainty regarding its effects, deployment must be suspended without delay in accordance with the precautionary principle.

Article 9 — Promotion of Innovations Serving Peace and Human Well-Being

States, institutions, and economic actors bear the responsibility of directing the development of artificial intelligence toward purposes beneficial to humanity.

They shall give priority support to uses contributing to the protection of life, the improvement of living conditions, access to food, healthcare, education, diversity, inclusion, cooperation among peoples, and the peaceful resolution of global crises.

They shall also promote capacity transfer, knowledge sharing, and equitable access to technologies in order to prevent technological divides between States and among populations.

Article 10 — Environment and Digital Sobriety

The energy, carbon, and water footprint of artificial intelligence systems constitutes a real externality that must be measured, reduced, and, where appropriate, made public.

The computing power employed must be proportionate to the social utility pursued.

No large-scale deployment should occur without a prior assessment of its environmental impact.

The principle of digital sobriety must guide the design, training, and use of artificial intelligence systems.

Article 11 — Democratic Governance and Accountability

The governance of artificial intelligence must be founded on transparency, participation, competence, pluralism, and accountability.

Major decisions concerning its development and uses must involve international organizations, States, economic actors, the research community, civil society, and citizens.

A clear framework of responsibility must make it possible to identify the actors responsible at each stage of the systems’ life cycle and to guarantee effective avenues for remedy and redress.

An annexed protocol may specify the composition, missions, and functioning of an international oversight body, in coordination with work carried out within the United Nations.

Final Clause — Common Commitments

The signatories acknowledge their collective responsibility in guiding the development of artificial intelligence.

They undertake to make it an instrument serving life, freedom, peace, and shared progress; to guarantee the maintenance of human control; and to act without delay against any drift liable to threaten humanity, legitimate institutions, or future generations.